mirror of
https://github.com/ziglang/zig.git
synced 2024-12-04 19:09:32 +00:00
4180 lines
112 KiB
Zig
4180 lines
112 KiB
Zig
//! This file is auto-generated by tools/update_cpu_features.zig.
|
|
|
|
const std = @import("../std.zig");
|
|
const CpuFeature = std.Target.Cpu.Feature;
|
|
const CpuModel = std.Target.Cpu.Model;
|
|
|
|
pub const Feature = enum {
|
|
@"16bit_mode",
|
|
@"32bit_mode",
|
|
@"3dnow",
|
|
@"3dnowa",
|
|
@"64bit",
|
|
adx,
|
|
aes,
|
|
allow_light_256_bit,
|
|
amx_bf16,
|
|
amx_complex,
|
|
amx_fp16,
|
|
amx_int8,
|
|
amx_tile,
|
|
avx,
|
|
avx2,
|
|
avx512bf16,
|
|
avx512bitalg,
|
|
avx512bw,
|
|
avx512cd,
|
|
avx512dq,
|
|
avx512er,
|
|
avx512f,
|
|
avx512fp16,
|
|
avx512ifma,
|
|
avx512pf,
|
|
avx512vbmi,
|
|
avx512vbmi2,
|
|
avx512vl,
|
|
avx512vnni,
|
|
avx512vp2intersect,
|
|
avx512vpopcntdq,
|
|
avxifma,
|
|
avxneconvert,
|
|
avxvnni,
|
|
avxvnniint16,
|
|
avxvnniint8,
|
|
bmi,
|
|
bmi2,
|
|
branchfusion,
|
|
cldemote,
|
|
clflushopt,
|
|
clwb,
|
|
clzero,
|
|
cmov,
|
|
cmpccxadd,
|
|
crc32,
|
|
cx16,
|
|
cx8,
|
|
enqcmd,
|
|
ermsb,
|
|
f16c,
|
|
false_deps_getmant,
|
|
false_deps_lzcnt_tzcnt,
|
|
false_deps_mulc,
|
|
false_deps_mullq,
|
|
false_deps_perm,
|
|
false_deps_popcnt,
|
|
false_deps_range,
|
|
fast_11bytenop,
|
|
fast_15bytenop,
|
|
fast_7bytenop,
|
|
fast_bextr,
|
|
fast_gather,
|
|
fast_hops,
|
|
fast_lzcnt,
|
|
fast_movbe,
|
|
fast_scalar_fsqrt,
|
|
fast_scalar_shift_masks,
|
|
fast_shld_rotate,
|
|
fast_variable_crosslane_shuffle,
|
|
fast_variable_perlane_shuffle,
|
|
fast_vector_fsqrt,
|
|
fast_vector_shift_masks,
|
|
faster_shift_than_shuffle,
|
|
fma,
|
|
fma4,
|
|
fsgsbase,
|
|
fsrm,
|
|
fxsr,
|
|
gfni,
|
|
harden_sls_ijmp,
|
|
harden_sls_ret,
|
|
hreset,
|
|
idivl_to_divb,
|
|
idivq_to_divl,
|
|
invpcid,
|
|
kl,
|
|
lea_sp,
|
|
lea_uses_ag,
|
|
lvi_cfi,
|
|
lvi_load_hardening,
|
|
lwp,
|
|
lzcnt,
|
|
macrofusion,
|
|
mmx,
|
|
movbe,
|
|
movdir64b,
|
|
movdiri,
|
|
mwaitx,
|
|
no_bypass_delay,
|
|
no_bypass_delay_blend,
|
|
no_bypass_delay_mov,
|
|
no_bypass_delay_shuffle,
|
|
nopl,
|
|
pad_short_functions,
|
|
pclmul,
|
|
pconfig,
|
|
pku,
|
|
popcnt,
|
|
prefer_128_bit,
|
|
prefer_256_bit,
|
|
prefer_mask_registers,
|
|
prefer_movmsk_over_vtest,
|
|
prefetchi,
|
|
prefetchwt1,
|
|
prfchw,
|
|
ptwrite,
|
|
raoint,
|
|
rdpid,
|
|
rdpru,
|
|
rdrnd,
|
|
rdseed,
|
|
retpoline,
|
|
retpoline_external_thunk,
|
|
retpoline_indirect_branches,
|
|
retpoline_indirect_calls,
|
|
rtm,
|
|
sahf,
|
|
sbb_dep_breaking,
|
|
serialize,
|
|
seses,
|
|
sgx,
|
|
sha,
|
|
sha512,
|
|
shstk,
|
|
slow_3ops_lea,
|
|
slow_incdec,
|
|
slow_lea,
|
|
slow_pmaddwd,
|
|
slow_pmulld,
|
|
slow_shld,
|
|
slow_two_mem_ops,
|
|
slow_unaligned_mem_16,
|
|
slow_unaligned_mem_32,
|
|
sm3,
|
|
sm4,
|
|
soft_float,
|
|
sse,
|
|
sse2,
|
|
sse3,
|
|
sse4_1,
|
|
sse4_2,
|
|
sse4a,
|
|
sse_unaligned_mem,
|
|
ssse3,
|
|
tagged_globals,
|
|
tbm,
|
|
tsxldtrk,
|
|
tuning_fast_imm_vector_shift,
|
|
uintr,
|
|
use_glm_div_sqrt_costs,
|
|
use_slm_arith_costs,
|
|
vaes,
|
|
vpclmulqdq,
|
|
vzeroupper,
|
|
waitpkg,
|
|
wbnoinvd,
|
|
widekl,
|
|
x87,
|
|
xop,
|
|
xsave,
|
|
xsavec,
|
|
xsaveopt,
|
|
xsaves,
|
|
};
|
|
|
|
pub const featureSet = CpuFeature.feature_set_fns(Feature).featureSet;
|
|
pub const featureSetHas = CpuFeature.feature_set_fns(Feature).featureSetHas;
|
|
pub const featureSetHasAny = CpuFeature.feature_set_fns(Feature).featureSetHasAny;
|
|
pub const featureSetHasAll = CpuFeature.feature_set_fns(Feature).featureSetHasAll;
|
|
|
|
pub const all_features = blk: {
|
|
const len = @typeInfo(Feature).Enum.fields.len;
|
|
std.debug.assert(len <= CpuFeature.Set.needed_bit_count);
|
|
var result: [len]CpuFeature = undefined;
|
|
result[@intFromEnum(Feature.@"16bit_mode")] = .{
|
|
.llvm_name = "16bit-mode",
|
|
.description = "16-bit mode (i8086)",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.@"32bit_mode")] = .{
|
|
.llvm_name = "32bit-mode",
|
|
.description = "32-bit mode (80386)",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.@"3dnow")] = .{
|
|
.llvm_name = "3dnow",
|
|
.description = "Enable 3DNow! instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.mmx,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.@"3dnowa")] = .{
|
|
.llvm_name = "3dnowa",
|
|
.description = "Enable 3DNow! Athlon instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.@"3dnow",
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.@"64bit")] = .{
|
|
.llvm_name = "64bit",
|
|
.description = "Support 64-bit instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.adx)] = .{
|
|
.llvm_name = "adx",
|
|
.description = "Support ADX instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.aes)] = .{
|
|
.llvm_name = "aes",
|
|
.description = "Enable AES instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.sse2,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.allow_light_256_bit)] = .{
|
|
.llvm_name = "allow-light-256-bit",
|
|
.description = "Enable generation of 256-bit load/stores even if we prefer 128-bit",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.amx_bf16)] = .{
|
|
.llvm_name = "amx-bf16",
|
|
.description = "Support AMX-BF16 instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.amx_tile,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.amx_complex)] = .{
|
|
.llvm_name = "amx-complex",
|
|
.description = "Support AMX-COMPLEX instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.amx_tile,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.amx_fp16)] = .{
|
|
.llvm_name = "amx-fp16",
|
|
.description = "Support AMX amx-fp16 instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.amx_tile,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.amx_int8)] = .{
|
|
.llvm_name = "amx-int8",
|
|
.description = "Support AMX-INT8 instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.amx_tile,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.amx_tile)] = .{
|
|
.llvm_name = "amx-tile",
|
|
.description = "Support AMX-TILE instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.avx)] = .{
|
|
.llvm_name = "avx",
|
|
.description = "Enable AVX instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.sse4_2,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx2)] = .{
|
|
.llvm_name = "avx2",
|
|
.description = "Enable AVX2 instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx512bf16)] = .{
|
|
.llvm_name = "avx512bf16",
|
|
.description = "Support bfloat16 floating point",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx512bw,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx512bitalg)] = .{
|
|
.llvm_name = "avx512bitalg",
|
|
.description = "Enable AVX-512 Bit Algorithms",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx512bw,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx512bw)] = .{
|
|
.llvm_name = "avx512bw",
|
|
.description = "Enable AVX-512 Byte and Word Instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx512f,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx512cd)] = .{
|
|
.llvm_name = "avx512cd",
|
|
.description = "Enable AVX-512 Conflict Detection Instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx512f,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx512dq)] = .{
|
|
.llvm_name = "avx512dq",
|
|
.description = "Enable AVX-512 Doubleword and Quadword Instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx512f,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx512er)] = .{
|
|
.llvm_name = "avx512er",
|
|
.description = "Enable AVX-512 Exponential and Reciprocal Instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx512f,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx512f)] = .{
|
|
.llvm_name = "avx512f",
|
|
.description = "Enable AVX-512 instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx2,
|
|
.f16c,
|
|
.fma,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx512fp16)] = .{
|
|
.llvm_name = "avx512fp16",
|
|
.description = "Support 16-bit floating point",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx512bw,
|
|
.avx512dq,
|
|
.avx512vl,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx512ifma)] = .{
|
|
.llvm_name = "avx512ifma",
|
|
.description = "Enable AVX-512 Integer Fused Multiple-Add",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx512f,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx512pf)] = .{
|
|
.llvm_name = "avx512pf",
|
|
.description = "Enable AVX-512 PreFetch Instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx512f,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx512vbmi)] = .{
|
|
.llvm_name = "avx512vbmi",
|
|
.description = "Enable AVX-512 Vector Byte Manipulation Instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx512bw,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx512vbmi2)] = .{
|
|
.llvm_name = "avx512vbmi2",
|
|
.description = "Enable AVX-512 further Vector Byte Manipulation Instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx512bw,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx512vl)] = .{
|
|
.llvm_name = "avx512vl",
|
|
.description = "Enable AVX-512 Vector Length eXtensions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx512f,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx512vnni)] = .{
|
|
.llvm_name = "avx512vnni",
|
|
.description = "Enable AVX-512 Vector Neural Network Instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx512f,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx512vp2intersect)] = .{
|
|
.llvm_name = "avx512vp2intersect",
|
|
.description = "Enable AVX-512 vp2intersect",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx512f,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avx512vpopcntdq)] = .{
|
|
.llvm_name = "avx512vpopcntdq",
|
|
.description = "Enable AVX-512 Population Count Instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx512f,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avxifma)] = .{
|
|
.llvm_name = "avxifma",
|
|
.description = "Enable AVX-IFMA",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx2,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avxneconvert)] = .{
|
|
.llvm_name = "avxneconvert",
|
|
.description = "Support AVX-NE-CONVERT instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx2,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avxvnni)] = .{
|
|
.llvm_name = "avxvnni",
|
|
.description = "Support AVX_VNNI encoding",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx2,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avxvnniint16)] = .{
|
|
.llvm_name = "avxvnniint16",
|
|
.description = "Enable AVX-VNNI-INT16",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx2,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.avxvnniint8)] = .{
|
|
.llvm_name = "avxvnniint8",
|
|
.description = "Enable AVX-VNNI-INT8",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx2,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.bmi)] = .{
|
|
.llvm_name = "bmi",
|
|
.description = "Support BMI instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.bmi2)] = .{
|
|
.llvm_name = "bmi2",
|
|
.description = "Support BMI2 instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.branchfusion)] = .{
|
|
.llvm_name = "branchfusion",
|
|
.description = "CMP/TEST can be fused with conditional branches",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.cldemote)] = .{
|
|
.llvm_name = "cldemote",
|
|
.description = "Enable Cache Line Demote",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.clflushopt)] = .{
|
|
.llvm_name = "clflushopt",
|
|
.description = "Flush A Cache Line Optimized",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.clwb)] = .{
|
|
.llvm_name = "clwb",
|
|
.description = "Cache Line Write Back",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.clzero)] = .{
|
|
.llvm_name = "clzero",
|
|
.description = "Enable Cache Line Zero",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.cmov)] = .{
|
|
.llvm_name = "cmov",
|
|
.description = "Enable conditional move instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.cmpccxadd)] = .{
|
|
.llvm_name = "cmpccxadd",
|
|
.description = "Support CMPCCXADD instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.crc32)] = .{
|
|
.llvm_name = "crc32",
|
|
.description = "Enable SSE 4.2 CRC32 instruction (used when SSE4.2 is supported but function is GPR only)",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.cx16)] = .{
|
|
.llvm_name = "cx16",
|
|
.description = "64-bit with cmpxchg16b (this is true for most x86-64 chips, but not the first AMD chips)",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.cx8,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.cx8)] = .{
|
|
.llvm_name = "cx8",
|
|
.description = "Support CMPXCHG8B instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.enqcmd)] = .{
|
|
.llvm_name = "enqcmd",
|
|
.description = "Has ENQCMD instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.ermsb)] = .{
|
|
.llvm_name = "ermsb",
|
|
.description = "REP MOVS/STOS are fast",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.f16c)] = .{
|
|
.llvm_name = "f16c",
|
|
.description = "Support 16-bit floating point conversion instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.false_deps_getmant)] = .{
|
|
.llvm_name = "false-deps-getmant",
|
|
.description = "VGETMANTSS/SD/SH and VGETMANDPS/PD(memory version) has a false dependency on dest register",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.false_deps_lzcnt_tzcnt)] = .{
|
|
.llvm_name = "false-deps-lzcnt-tzcnt",
|
|
.description = "LZCNT/TZCNT have a false dependency on dest register",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.false_deps_mulc)] = .{
|
|
.llvm_name = "false-deps-mulc",
|
|
.description = "VF[C]MULCPH/SH has a false dependency on dest register",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.false_deps_mullq)] = .{
|
|
.llvm_name = "false-deps-mullq",
|
|
.description = "VPMULLQ has a false dependency on dest register",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.false_deps_perm)] = .{
|
|
.llvm_name = "false-deps-perm",
|
|
.description = "VPERMD/Q/PS/PD has a false dependency on dest register",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.false_deps_popcnt)] = .{
|
|
.llvm_name = "false-deps-popcnt",
|
|
.description = "POPCNT has a false dependency on dest register",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.false_deps_range)] = .{
|
|
.llvm_name = "false-deps-range",
|
|
.description = "VRANGEPD/PS/SD/SS has a false dependency on dest register",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fast_11bytenop)] = .{
|
|
.llvm_name = "fast-11bytenop",
|
|
.description = "Target can quickly decode up to 11 byte NOPs",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fast_15bytenop)] = .{
|
|
.llvm_name = "fast-15bytenop",
|
|
.description = "Target can quickly decode up to 15 byte NOPs",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fast_7bytenop)] = .{
|
|
.llvm_name = "fast-7bytenop",
|
|
.description = "Target can quickly decode up to 7 byte NOPs",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fast_bextr)] = .{
|
|
.llvm_name = "fast-bextr",
|
|
.description = "Indicates that the BEXTR instruction is implemented as a single uop with good throughput",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fast_gather)] = .{
|
|
.llvm_name = "fast-gather",
|
|
.description = "Indicates if gather is reasonably fast (this is true for Skylake client and all AVX-512 CPUs)",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fast_hops)] = .{
|
|
.llvm_name = "fast-hops",
|
|
.description = "Prefer horizontal vector math instructions (haddp, phsub, etc.) over normal vector instructions with shuffles",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fast_lzcnt)] = .{
|
|
.llvm_name = "fast-lzcnt",
|
|
.description = "LZCNT instructions are as fast as most simple integer ops",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fast_movbe)] = .{
|
|
.llvm_name = "fast-movbe",
|
|
.description = "Prefer a movbe over a single-use load + bswap / single-use bswap + store",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fast_scalar_fsqrt)] = .{
|
|
.llvm_name = "fast-scalar-fsqrt",
|
|
.description = "Scalar SQRT is fast (disable Newton-Raphson)",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fast_scalar_shift_masks)] = .{
|
|
.llvm_name = "fast-scalar-shift-masks",
|
|
.description = "Prefer a left/right scalar logical shift pair over a shift+and pair",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fast_shld_rotate)] = .{
|
|
.llvm_name = "fast-shld-rotate",
|
|
.description = "SHLD can be used as a faster rotate",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fast_variable_crosslane_shuffle)] = .{
|
|
.llvm_name = "fast-variable-crosslane-shuffle",
|
|
.description = "Cross-lane shuffles with variable masks are fast",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fast_variable_perlane_shuffle)] = .{
|
|
.llvm_name = "fast-variable-perlane-shuffle",
|
|
.description = "Per-lane shuffles with variable masks are fast",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fast_vector_fsqrt)] = .{
|
|
.llvm_name = "fast-vector-fsqrt",
|
|
.description = "Vector SQRT is fast (disable Newton-Raphson)",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fast_vector_shift_masks)] = .{
|
|
.llvm_name = "fast-vector-shift-masks",
|
|
.description = "Prefer a left/right vector logical shift pair over a shift+and pair",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.faster_shift_than_shuffle)] = .{
|
|
.llvm_name = "faster-shift-than-shuffle",
|
|
.description = "Shifts are faster (or as fast) as shuffle",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fma)] = .{
|
|
.llvm_name = "fma",
|
|
.description = "Enable three-operand fused multiple-add",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.fma4)] = .{
|
|
.llvm_name = "fma4",
|
|
.description = "Enable four-operand fused multiple-add",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx,
|
|
.sse4a,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.fsgsbase)] = .{
|
|
.llvm_name = "fsgsbase",
|
|
.description = "Support FS/GS Base instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fsrm)] = .{
|
|
.llvm_name = "fsrm",
|
|
.description = "REP MOVSB of short lengths is faster",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.fxsr)] = .{
|
|
.llvm_name = "fxsr",
|
|
.description = "Support fxsave/fxrestore instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.gfni)] = .{
|
|
.llvm_name = "gfni",
|
|
.description = "Enable Galois Field Arithmetic Instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.sse2,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.harden_sls_ijmp)] = .{
|
|
.llvm_name = "harden-sls-ijmp",
|
|
.description = "Harden against straight line speculation across indirect JMP instructions.",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.harden_sls_ret)] = .{
|
|
.llvm_name = "harden-sls-ret",
|
|
.description = "Harden against straight line speculation across RET instructions.",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.hreset)] = .{
|
|
.llvm_name = "hreset",
|
|
.description = "Has hreset instruction",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.idivl_to_divb)] = .{
|
|
.llvm_name = "idivl-to-divb",
|
|
.description = "Use 8-bit divide for positive values less than 256",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.idivq_to_divl)] = .{
|
|
.llvm_name = "idivq-to-divl",
|
|
.description = "Use 32-bit divide for positive values less than 2^32",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.invpcid)] = .{
|
|
.llvm_name = "invpcid",
|
|
.description = "Invalidate Process-Context Identifier",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.kl)] = .{
|
|
.llvm_name = "kl",
|
|
.description = "Support Key Locker kl Instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.sse2,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.lea_sp)] = .{
|
|
.llvm_name = "lea-sp",
|
|
.description = "Use LEA for adjusting the stack pointer (this is an optimization for Intel Atom processors)",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.lea_uses_ag)] = .{
|
|
.llvm_name = "lea-uses-ag",
|
|
.description = "LEA instruction needs inputs at AG stage",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.lvi_cfi)] = .{
|
|
.llvm_name = "lvi-cfi",
|
|
.description = "Prevent indirect calls/branches from using a memory operand, and precede all indirect calls/branches from a register with an LFENCE instruction to serialize control flow. Also decompose RET instructions into a POP+LFENCE+JMP sequence.",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.lvi_load_hardening)] = .{
|
|
.llvm_name = "lvi-load-hardening",
|
|
.description = "Insert LFENCE instructions to prevent data speculatively injected into loads from being used maliciously.",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.lwp)] = .{
|
|
.llvm_name = "lwp",
|
|
.description = "Enable LWP instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.lzcnt)] = .{
|
|
.llvm_name = "lzcnt",
|
|
.description = "Support LZCNT instruction",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.macrofusion)] = .{
|
|
.llvm_name = "macrofusion",
|
|
.description = "Various instructions can be fused with conditional branches",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.mmx)] = .{
|
|
.llvm_name = "mmx",
|
|
.description = "Enable MMX instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.movbe)] = .{
|
|
.llvm_name = "movbe",
|
|
.description = "Support MOVBE instruction",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.movdir64b)] = .{
|
|
.llvm_name = "movdir64b",
|
|
.description = "Support movdir64b instruction (direct store 64 bytes)",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.movdiri)] = .{
|
|
.llvm_name = "movdiri",
|
|
.description = "Support movdiri instruction (direct store integer)",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.mwaitx)] = .{
|
|
.llvm_name = "mwaitx",
|
|
.description = "Enable MONITORX/MWAITX timer functionality",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.no_bypass_delay)] = .{
|
|
.llvm_name = "no-bypass-delay",
|
|
.description = "Has no bypass delay when using the 'wrong' domain",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.no_bypass_delay_blend)] = .{
|
|
.llvm_name = "no-bypass-delay-blend",
|
|
.description = "Has no bypass delay when using the 'wrong' blend type",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.no_bypass_delay_mov)] = .{
|
|
.llvm_name = "no-bypass-delay-mov",
|
|
.description = "Has no bypass delay when using the 'wrong' mov type",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.no_bypass_delay_shuffle)] = .{
|
|
.llvm_name = "no-bypass-delay-shuffle",
|
|
.description = "Has no bypass delay when using the 'wrong' shuffle type",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.nopl)] = .{
|
|
.llvm_name = "nopl",
|
|
.description = "Enable NOPL instruction (generally pentium pro+)",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.pad_short_functions)] = .{
|
|
.llvm_name = "pad-short-functions",
|
|
.description = "Pad short functions (to prevent a stall when returning too early)",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.pclmul)] = .{
|
|
.llvm_name = "pclmul",
|
|
.description = "Enable packed carry-less multiplication instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.sse2,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.pconfig)] = .{
|
|
.llvm_name = "pconfig",
|
|
.description = "platform configuration instruction",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.pku)] = .{
|
|
.llvm_name = "pku",
|
|
.description = "Enable protection keys",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.popcnt)] = .{
|
|
.llvm_name = "popcnt",
|
|
.description = "Support POPCNT instruction",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.prefer_128_bit)] = .{
|
|
.llvm_name = "prefer-128-bit",
|
|
.description = "Prefer 128-bit AVX instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.prefer_256_bit)] = .{
|
|
.llvm_name = "prefer-256-bit",
|
|
.description = "Prefer 256-bit AVX instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.prefer_mask_registers)] = .{
|
|
.llvm_name = "prefer-mask-registers",
|
|
.description = "Prefer AVX512 mask registers over PTEST/MOVMSK",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.prefer_movmsk_over_vtest)] = .{
|
|
.llvm_name = "prefer-movmsk-over-vtest",
|
|
.description = "Prefer movmsk over vtest instruction",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.prefetchi)] = .{
|
|
.llvm_name = "prefetchi",
|
|
.description = "Prefetch instruction with T0 or T1 Hint",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.prefetchwt1)] = .{
|
|
.llvm_name = "prefetchwt1",
|
|
.description = "Prefetch with Intent to Write and T1 Hint",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.prfchw)] = .{
|
|
.llvm_name = "prfchw",
|
|
.description = "Support PRFCHW instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.ptwrite)] = .{
|
|
.llvm_name = "ptwrite",
|
|
.description = "Support ptwrite instruction",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.raoint)] = .{
|
|
.llvm_name = "raoint",
|
|
.description = "Support RAO-INT instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.rdpid)] = .{
|
|
.llvm_name = "rdpid",
|
|
.description = "Support RDPID instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.rdpru)] = .{
|
|
.llvm_name = "rdpru",
|
|
.description = "Support RDPRU instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.rdrnd)] = .{
|
|
.llvm_name = "rdrnd",
|
|
.description = "Support RDRAND instruction",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.rdseed)] = .{
|
|
.llvm_name = "rdseed",
|
|
.description = "Support RDSEED instruction",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.retpoline)] = .{
|
|
.llvm_name = "retpoline",
|
|
.description = "Remove speculation of indirect branches from the generated code, either by avoiding them entirely or lowering them with a speculation blocking construct",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.retpoline_indirect_branches,
|
|
.retpoline_indirect_calls,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.retpoline_external_thunk)] = .{
|
|
.llvm_name = "retpoline-external-thunk",
|
|
.description = "When lowering an indirect call or branch using a `retpoline`, rely on the specified user provided thunk rather than emitting one ourselves. Only has effect when combined with some other retpoline feature",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.retpoline_indirect_calls,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.retpoline_indirect_branches)] = .{
|
|
.llvm_name = "retpoline-indirect-branches",
|
|
.description = "Remove speculation of indirect branches from the generated code",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.retpoline_indirect_calls)] = .{
|
|
.llvm_name = "retpoline-indirect-calls",
|
|
.description = "Remove speculation of indirect calls from the generated code",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.rtm)] = .{
|
|
.llvm_name = "rtm",
|
|
.description = "Support RTM instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.sahf)] = .{
|
|
.llvm_name = "sahf",
|
|
.description = "Support LAHF and SAHF instructions in 64-bit mode",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.sbb_dep_breaking)] = .{
|
|
.llvm_name = "sbb-dep-breaking",
|
|
.description = "SBB with same register has no source dependency",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.serialize)] = .{
|
|
.llvm_name = "serialize",
|
|
.description = "Has serialize instruction",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.seses)] = .{
|
|
.llvm_name = "seses",
|
|
.description = "Prevent speculative execution side channel timing attacks by inserting a speculation barrier before memory reads, memory writes, and conditional branches. Implies LVI Control Flow integrity.",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.lvi_cfi,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.sgx)] = .{
|
|
.llvm_name = "sgx",
|
|
.description = "Enable Software Guard Extensions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.sha)] = .{
|
|
.llvm_name = "sha",
|
|
.description = "Enable SHA instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.sse2,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.sha512)] = .{
|
|
.llvm_name = "sha512",
|
|
.description = "Support SHA512 instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.shstk)] = .{
|
|
.llvm_name = "shstk",
|
|
.description = "Support CET Shadow-Stack instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.slow_3ops_lea)] = .{
|
|
.llvm_name = "slow-3ops-lea",
|
|
.description = "LEA instruction with 3 ops or certain registers is slow",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.slow_incdec)] = .{
|
|
.llvm_name = "slow-incdec",
|
|
.description = "INC and DEC instructions are slower than ADD and SUB",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.slow_lea)] = .{
|
|
.llvm_name = "slow-lea",
|
|
.description = "LEA instruction with certain arguments is slow",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.slow_pmaddwd)] = .{
|
|
.llvm_name = "slow-pmaddwd",
|
|
.description = "PMADDWD is slower than PMULLD",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.slow_pmulld)] = .{
|
|
.llvm_name = "slow-pmulld",
|
|
.description = "PMULLD instruction is slow (compared to PMULLW/PMULHW and PMULUDQ)",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.slow_shld)] = .{
|
|
.llvm_name = "slow-shld",
|
|
.description = "SHLD instruction is slow",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.slow_two_mem_ops)] = .{
|
|
.llvm_name = "slow-two-mem-ops",
|
|
.description = "Two memory operand instructions are slow",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.slow_unaligned_mem_16)] = .{
|
|
.llvm_name = "slow-unaligned-mem-16",
|
|
.description = "Slow unaligned 16-byte memory access",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.slow_unaligned_mem_32)] = .{
|
|
.llvm_name = "slow-unaligned-mem-32",
|
|
.description = "Slow unaligned 32-byte memory access",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.sm3)] = .{
|
|
.llvm_name = "sm3",
|
|
.description = "Support SM3 instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.sm4)] = .{
|
|
.llvm_name = "sm4",
|
|
.description = "Support SM4 instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.soft_float)] = .{
|
|
.llvm_name = "soft-float",
|
|
.description = "Use software floating point features",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.sse)] = .{
|
|
.llvm_name = "sse",
|
|
.description = "Enable SSE instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.sse2)] = .{
|
|
.llvm_name = "sse2",
|
|
.description = "Enable SSE2 instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.sse,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.sse3)] = .{
|
|
.llvm_name = "sse3",
|
|
.description = "Enable SSE3 instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.sse2,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.sse4_1)] = .{
|
|
.llvm_name = "sse4.1",
|
|
.description = "Enable SSE 4.1 instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.ssse3,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.sse4_2)] = .{
|
|
.llvm_name = "sse4.2",
|
|
.description = "Enable SSE 4.2 instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.sse4_1,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.sse4a)] = .{
|
|
.llvm_name = "sse4a",
|
|
.description = "Support SSE 4a instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.sse3,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.sse_unaligned_mem)] = .{
|
|
.llvm_name = "sse-unaligned-mem",
|
|
.description = "Allow unaligned memory operands with SSE instructions (this may require setting a configuration bit in the processor)",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.ssse3)] = .{
|
|
.llvm_name = "ssse3",
|
|
.description = "Enable SSSE3 instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.sse3,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.tagged_globals)] = .{
|
|
.llvm_name = "tagged-globals",
|
|
.description = "Use an instruction sequence for taking the address of a global that allows a memory tag in the upper address bits.",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.tbm)] = .{
|
|
.llvm_name = "tbm",
|
|
.description = "Enable TBM instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.tsxldtrk)] = .{
|
|
.llvm_name = "tsxldtrk",
|
|
.description = "Support TSXLDTRK instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.tuning_fast_imm_vector_shift)] = .{
|
|
.llvm_name = "tuning-fast-imm-vector-shift",
|
|
.description = "Vector shifts are fast (2/cycle) as opposed to slow (1/cycle)",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.uintr)] = .{
|
|
.llvm_name = "uintr",
|
|
.description = "Has UINTR Instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.use_glm_div_sqrt_costs)] = .{
|
|
.llvm_name = "use-glm-div-sqrt-costs",
|
|
.description = "Use Goldmont specific floating point div/sqrt costs",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.use_slm_arith_costs)] = .{
|
|
.llvm_name = "use-slm-arith-costs",
|
|
.description = "Use Silvermont specific arithmetic costs",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.vaes)] = .{
|
|
.llvm_name = "vaes",
|
|
.description = "Promote selected AES instructions to AVX512/AVX registers",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.aes,
|
|
.avx,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.vpclmulqdq)] = .{
|
|
.llvm_name = "vpclmulqdq",
|
|
.description = "Enable vpclmulqdq instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.avx,
|
|
.pclmul,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.vzeroupper)] = .{
|
|
.llvm_name = "vzeroupper",
|
|
.description = "Should insert vzeroupper instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.waitpkg)] = .{
|
|
.llvm_name = "waitpkg",
|
|
.description = "Wait and pause enhancements",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.wbnoinvd)] = .{
|
|
.llvm_name = "wbnoinvd",
|
|
.description = "Write Back No Invalidate",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.widekl)] = .{
|
|
.llvm_name = "widekl",
|
|
.description = "Support Key Locker wide Instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.kl,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.x87)] = .{
|
|
.llvm_name = "x87",
|
|
.description = "Enable X87 float instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.xop)] = .{
|
|
.llvm_name = "xop",
|
|
.description = "Enable XOP instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.fma4,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.xsave)] = .{
|
|
.llvm_name = "xsave",
|
|
.description = "Support xsave instructions",
|
|
.dependencies = featureSet(&[_]Feature{}),
|
|
};
|
|
result[@intFromEnum(Feature.xsavec)] = .{
|
|
.llvm_name = "xsavec",
|
|
.description = "Support xsavec instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.xsave,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.xsaveopt)] = .{
|
|
.llvm_name = "xsaveopt",
|
|
.description = "Support xsaveopt instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.xsave,
|
|
}),
|
|
};
|
|
result[@intFromEnum(Feature.xsaves)] = .{
|
|
.llvm_name = "xsaves",
|
|
.description = "Support xsaves instructions",
|
|
.dependencies = featureSet(&[_]Feature{
|
|
.xsave,
|
|
}),
|
|
};
|
|
const ti = @typeInfo(Feature);
|
|
for (&result, 0..) |*elem, i| {
|
|
elem.index = i;
|
|
elem.name = ti.Enum.fields[i].name;
|
|
}
|
|
break :blk result;
|
|
};
|
|
|
|
pub const cpu = struct {
|
|
pub const alderlake = CpuModel{
|
|
.name = "alderlake",
|
|
.llvm_name = "alderlake",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.allow_light_256_bit,
|
|
.avxvnni,
|
|
.bmi,
|
|
.bmi2,
|
|
.cldemote,
|
|
.clflushopt,
|
|
.clwb,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.f16c,
|
|
.false_deps_perm,
|
|
.false_deps_popcnt,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fma,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.gfni,
|
|
.hreset,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.movdir64b,
|
|
.movdiri,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pconfig,
|
|
.pku,
|
|
.popcnt,
|
|
.prefer_movmsk_over_vtest,
|
|
.prfchw,
|
|
.ptwrite,
|
|
.rdpid,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.serialize,
|
|
.sha,
|
|
.shstk,
|
|
.slow_3ops_lea,
|
|
.tuning_fast_imm_vector_shift,
|
|
.vaes,
|
|
.vpclmulqdq,
|
|
.vzeroupper,
|
|
.waitpkg,
|
|
.widekl,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const amdfam10 = CpuModel{
|
|
.name = "amdfam10",
|
|
.llvm_name = "amdfam10",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnowa",
|
|
.@"64bit",
|
|
.cmov,
|
|
.cx16,
|
|
.fast_scalar_shift_masks,
|
|
.fxsr,
|
|
.lzcnt,
|
|
.nopl,
|
|
.popcnt,
|
|
.prfchw,
|
|
.sahf,
|
|
.sbb_dep_breaking,
|
|
.slow_shld,
|
|
.sse4a,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const athlon = CpuModel{
|
|
.name = "athlon",
|
|
.llvm_name = "athlon",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnowa",
|
|
.cmov,
|
|
.cx8,
|
|
.nopl,
|
|
.slow_shld,
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const athlon64 = CpuModel{
|
|
.name = "athlon64",
|
|
.llvm_name = "athlon64",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnowa",
|
|
.@"64bit",
|
|
.cmov,
|
|
.cx8,
|
|
.fast_scalar_shift_masks,
|
|
.fxsr,
|
|
.nopl,
|
|
.sbb_dep_breaking,
|
|
.slow_shld,
|
|
.slow_unaligned_mem_16,
|
|
.sse2,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const athlon64_sse3 = CpuModel{
|
|
.name = "athlon64_sse3",
|
|
.llvm_name = "athlon64-sse3",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnowa",
|
|
.@"64bit",
|
|
.cmov,
|
|
.cx16,
|
|
.fast_scalar_shift_masks,
|
|
.fxsr,
|
|
.nopl,
|
|
.sbb_dep_breaking,
|
|
.slow_shld,
|
|
.slow_unaligned_mem_16,
|
|
.sse3,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const athlon_4 = CpuModel{
|
|
.name = "athlon_4",
|
|
.llvm_name = "athlon-4",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnowa",
|
|
.cmov,
|
|
.cx8,
|
|
.fxsr,
|
|
.nopl,
|
|
.slow_shld,
|
|
.slow_unaligned_mem_16,
|
|
.sse,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const athlon_fx = CpuModel{
|
|
.name = "athlon_fx",
|
|
.llvm_name = "athlon-fx",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnowa",
|
|
.@"64bit",
|
|
.cmov,
|
|
.cx8,
|
|
.fast_scalar_shift_masks,
|
|
.fxsr,
|
|
.nopl,
|
|
.sbb_dep_breaking,
|
|
.slow_shld,
|
|
.slow_unaligned_mem_16,
|
|
.sse2,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const athlon_mp = CpuModel{
|
|
.name = "athlon_mp",
|
|
.llvm_name = "athlon-mp",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnowa",
|
|
.cmov,
|
|
.cx8,
|
|
.fxsr,
|
|
.nopl,
|
|
.slow_shld,
|
|
.slow_unaligned_mem_16,
|
|
.sse,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const athlon_tbird = CpuModel{
|
|
.name = "athlon_tbird",
|
|
.llvm_name = "athlon-tbird",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnowa",
|
|
.cmov,
|
|
.cx8,
|
|
.nopl,
|
|
.slow_shld,
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const athlon_xp = CpuModel{
|
|
.name = "athlon_xp",
|
|
.llvm_name = "athlon-xp",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnowa",
|
|
.cmov,
|
|
.cx8,
|
|
.fxsr,
|
|
.nopl,
|
|
.slow_shld,
|
|
.slow_unaligned_mem_16,
|
|
.sse,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const atom = CpuModel{
|
|
.name = "atom",
|
|
.llvm_name = "atom",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.cmov,
|
|
.cx16,
|
|
.fxsr,
|
|
.idivl_to_divb,
|
|
.idivq_to_divl,
|
|
.lea_sp,
|
|
.lea_uses_ag,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay,
|
|
.nopl,
|
|
.pad_short_functions,
|
|
.sahf,
|
|
.slow_two_mem_ops,
|
|
.slow_unaligned_mem_16,
|
|
.ssse3,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const atom_sse4_2_movbe = CpuModel{
|
|
.name = "atom_sse4_2_movbe",
|
|
.llvm_name = "atom_sse4_2_movbe",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.aes,
|
|
.clflushopt,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.false_deps_popcnt,
|
|
.fast_7bytenop,
|
|
.fast_movbe,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prfchw,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.sha,
|
|
.slow_incdec,
|
|
.slow_lea,
|
|
.slow_pmulld,
|
|
.slow_two_mem_ops,
|
|
.sse4_2,
|
|
.use_slm_arith_costs,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const barcelona = CpuModel{
|
|
.name = "barcelona",
|
|
.llvm_name = "barcelona",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnowa",
|
|
.@"64bit",
|
|
.cmov,
|
|
.cx16,
|
|
.fast_scalar_shift_masks,
|
|
.fxsr,
|
|
.lzcnt,
|
|
.nopl,
|
|
.popcnt,
|
|
.prfchw,
|
|
.sahf,
|
|
.sbb_dep_breaking,
|
|
.slow_shld,
|
|
.sse4a,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const bdver1 = CpuModel{
|
|
.name = "bdver1",
|
|
.llvm_name = "bdver1",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.aes,
|
|
.branchfusion,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.fast_11bytenop,
|
|
.fast_scalar_shift_masks,
|
|
.fxsr,
|
|
.lwp,
|
|
.lzcnt,
|
|
.mmx,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prfchw,
|
|
.sahf,
|
|
.sbb_dep_breaking,
|
|
.slow_shld,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xop,
|
|
.xsave,
|
|
}),
|
|
};
|
|
pub const bdver2 = CpuModel{
|
|
.name = "bdver2",
|
|
.llvm_name = "bdver2",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.aes,
|
|
.bmi,
|
|
.branchfusion,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.f16c,
|
|
.fast_11bytenop,
|
|
.fast_bextr,
|
|
.fast_movbe,
|
|
.fast_scalar_shift_masks,
|
|
.fma,
|
|
.fxsr,
|
|
.lwp,
|
|
.lzcnt,
|
|
.mmx,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prfchw,
|
|
.sahf,
|
|
.sbb_dep_breaking,
|
|
.slow_shld,
|
|
.tbm,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xop,
|
|
.xsave,
|
|
}),
|
|
};
|
|
pub const bdver3 = CpuModel{
|
|
.name = "bdver3",
|
|
.llvm_name = "bdver3",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.aes,
|
|
.bmi,
|
|
.branchfusion,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.f16c,
|
|
.fast_11bytenop,
|
|
.fast_bextr,
|
|
.fast_movbe,
|
|
.fast_scalar_shift_masks,
|
|
.fma,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.lwp,
|
|
.lzcnt,
|
|
.mmx,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prfchw,
|
|
.sahf,
|
|
.sbb_dep_breaking,
|
|
.slow_shld,
|
|
.tbm,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xop,
|
|
.xsaveopt,
|
|
}),
|
|
};
|
|
pub const bdver4 = CpuModel{
|
|
.name = "bdver4",
|
|
.llvm_name = "bdver4",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.aes,
|
|
.avx2,
|
|
.bmi,
|
|
.bmi2,
|
|
.branchfusion,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.f16c,
|
|
.fast_11bytenop,
|
|
.fast_bextr,
|
|
.fast_movbe,
|
|
.fast_scalar_shift_masks,
|
|
.fma,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.lwp,
|
|
.lzcnt,
|
|
.mmx,
|
|
.movbe,
|
|
.mwaitx,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prfchw,
|
|
.rdrnd,
|
|
.sahf,
|
|
.sbb_dep_breaking,
|
|
.slow_shld,
|
|
.tbm,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xop,
|
|
.xsaveopt,
|
|
}),
|
|
};
|
|
pub const bonnell = CpuModel{
|
|
.name = "bonnell",
|
|
.llvm_name = "bonnell",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.cmov,
|
|
.cx16,
|
|
.fxsr,
|
|
.idivl_to_divb,
|
|
.idivq_to_divl,
|
|
.lea_sp,
|
|
.lea_uses_ag,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay,
|
|
.nopl,
|
|
.pad_short_functions,
|
|
.sahf,
|
|
.slow_two_mem_ops,
|
|
.slow_unaligned_mem_16,
|
|
.ssse3,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const broadwell = CpuModel{
|
|
.name = "broadwell",
|
|
.llvm_name = "broadwell",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.allow_light_256_bit,
|
|
.avx2,
|
|
.bmi,
|
|
.bmi2,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.ermsb,
|
|
.f16c,
|
|
.false_deps_lzcnt_tzcnt,
|
|
.false_deps_popcnt,
|
|
.fast_15bytenop,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fma,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prfchw,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.slow_3ops_lea,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsaveopt,
|
|
}),
|
|
};
|
|
pub const btver1 = CpuModel{
|
|
.name = "btver1",
|
|
.llvm_name = "btver1",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.cmov,
|
|
.cx16,
|
|
.fast_15bytenop,
|
|
.fast_scalar_shift_masks,
|
|
.fast_vector_shift_masks,
|
|
.fxsr,
|
|
.lzcnt,
|
|
.mmx,
|
|
.nopl,
|
|
.popcnt,
|
|
.prfchw,
|
|
.sahf,
|
|
.sbb_dep_breaking,
|
|
.slow_shld,
|
|
.sse4a,
|
|
.ssse3,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const btver2 = CpuModel{
|
|
.name = "btver2",
|
|
.llvm_name = "btver2",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.aes,
|
|
.bmi,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.f16c,
|
|
.fast_15bytenop,
|
|
.fast_bextr,
|
|
.fast_hops,
|
|
.fast_lzcnt,
|
|
.fast_movbe,
|
|
.fast_scalar_shift_masks,
|
|
.fast_vector_shift_masks,
|
|
.fxsr,
|
|
.lzcnt,
|
|
.mmx,
|
|
.movbe,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prfchw,
|
|
.sahf,
|
|
.sbb_dep_breaking,
|
|
.slow_shld,
|
|
.sse4a,
|
|
.x87,
|
|
.xsaveopt,
|
|
}),
|
|
};
|
|
pub const c3 = CpuModel{
|
|
.name = "c3",
|
|
.llvm_name = "c3",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnow",
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const c3_2 = CpuModel{
|
|
.name = "c3_2",
|
|
.llvm_name = "c3-2",
|
|
.features = featureSet(&[_]Feature{
|
|
.cmov,
|
|
.cx8,
|
|
.fxsr,
|
|
.mmx,
|
|
.slow_unaligned_mem_16,
|
|
.sse,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const cannonlake = CpuModel{
|
|
.name = "cannonlake",
|
|
.llvm_name = "cannonlake",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.aes,
|
|
.allow_light_256_bit,
|
|
.avx512cd,
|
|
.avx512dq,
|
|
.avx512ifma,
|
|
.avx512vbmi,
|
|
.avx512vl,
|
|
.bmi,
|
|
.bmi2,
|
|
.clflushopt,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.ermsb,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pclmul,
|
|
.pku,
|
|
.popcnt,
|
|
.prefer_256_bit,
|
|
.prfchw,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.sha,
|
|
.slow_3ops_lea,
|
|
.tuning_fast_imm_vector_shift,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const cascadelake = CpuModel{
|
|
.name = "cascadelake",
|
|
.llvm_name = "cascadelake",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.aes,
|
|
.allow_light_256_bit,
|
|
.avx512bw,
|
|
.avx512cd,
|
|
.avx512dq,
|
|
.avx512vl,
|
|
.avx512vnni,
|
|
.bmi,
|
|
.bmi2,
|
|
.clflushopt,
|
|
.clwb,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.ermsb,
|
|
.false_deps_popcnt,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.faster_shift_than_shuffle,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pclmul,
|
|
.pku,
|
|
.popcnt,
|
|
.prefer_256_bit,
|
|
.prfchw,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.slow_3ops_lea,
|
|
.tuning_fast_imm_vector_shift,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const cooperlake = CpuModel{
|
|
.name = "cooperlake",
|
|
.llvm_name = "cooperlake",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.aes,
|
|
.allow_light_256_bit,
|
|
.avx512bf16,
|
|
.avx512cd,
|
|
.avx512dq,
|
|
.avx512vl,
|
|
.avx512vnni,
|
|
.bmi,
|
|
.bmi2,
|
|
.clflushopt,
|
|
.clwb,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.ermsb,
|
|
.false_deps_popcnt,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.faster_shift_than_shuffle,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pclmul,
|
|
.pku,
|
|
.popcnt,
|
|
.prefer_256_bit,
|
|
.prfchw,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.slow_3ops_lea,
|
|
.tuning_fast_imm_vector_shift,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const core2 = CpuModel{
|
|
.name = "core2",
|
|
.llvm_name = "core2",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.cmov,
|
|
.cx16,
|
|
.fxsr,
|
|
.macrofusion,
|
|
.mmx,
|
|
.nopl,
|
|
.sahf,
|
|
.slow_unaligned_mem_16,
|
|
.ssse3,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const corei7 = CpuModel{
|
|
.name = "corei7",
|
|
.llvm_name = "corei7",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.fxsr,
|
|
.macrofusion,
|
|
.mmx,
|
|
.no_bypass_delay_mov,
|
|
.nopl,
|
|
.popcnt,
|
|
.sahf,
|
|
.sse4_2,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const emeraldrapids = CpuModel{
|
|
.name = "emeraldrapids",
|
|
.llvm_name = "emeraldrapids",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.allow_light_256_bit,
|
|
.amx_bf16,
|
|
.amx_int8,
|
|
.avx512bf16,
|
|
.avx512bitalg,
|
|
.avx512cd,
|
|
.avx512fp16,
|
|
.avx512ifma,
|
|
.avx512vbmi,
|
|
.avx512vbmi2,
|
|
.avx512vnni,
|
|
.avx512vpopcntdq,
|
|
.avxvnni,
|
|
.bmi,
|
|
.bmi2,
|
|
.cldemote,
|
|
.clflushopt,
|
|
.clwb,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.enqcmd,
|
|
.ermsb,
|
|
.false_deps_getmant,
|
|
.false_deps_mulc,
|
|
.false_deps_mullq,
|
|
.false_deps_perm,
|
|
.false_deps_range,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fsgsbase,
|
|
.fsrm,
|
|
.fxsr,
|
|
.gfni,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.movdir64b,
|
|
.movdiri,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pconfig,
|
|
.pku,
|
|
.popcnt,
|
|
.prefer_256_bit,
|
|
.prfchw,
|
|
.ptwrite,
|
|
.rdpid,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.serialize,
|
|
.sha,
|
|
.shstk,
|
|
.tsxldtrk,
|
|
.tuning_fast_imm_vector_shift,
|
|
.uintr,
|
|
.vaes,
|
|
.vpclmulqdq,
|
|
.vzeroupper,
|
|
.waitpkg,
|
|
.wbnoinvd,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const generic = CpuModel{
|
|
.name = "generic",
|
|
.llvm_name = "generic",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.cx8,
|
|
.fast_15bytenop,
|
|
.fast_scalar_fsqrt,
|
|
.idivq_to_divl,
|
|
.macrofusion,
|
|
.slow_3ops_lea,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const geode = CpuModel{
|
|
.name = "geode",
|
|
.llvm_name = "geode",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnowa",
|
|
.cx8,
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const goldmont = CpuModel{
|
|
.name = "goldmont",
|
|
.llvm_name = "goldmont",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.aes,
|
|
.clflushopt,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.false_deps_popcnt,
|
|
.fast_movbe,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prfchw,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.sha,
|
|
.slow_incdec,
|
|
.slow_lea,
|
|
.slow_two_mem_ops,
|
|
.sse4_2,
|
|
.use_glm_div_sqrt_costs,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const goldmont_plus = CpuModel{
|
|
.name = "goldmont_plus",
|
|
.llvm_name = "goldmont-plus",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.aes,
|
|
.clflushopt,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.fast_movbe,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prfchw,
|
|
.ptwrite,
|
|
.rdpid,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.sha,
|
|
.slow_incdec,
|
|
.slow_lea,
|
|
.slow_two_mem_ops,
|
|
.sse4_2,
|
|
.use_glm_div_sqrt_costs,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const grandridge = CpuModel{
|
|
.name = "grandridge",
|
|
.llvm_name = "grandridge",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.avxifma,
|
|
.avxneconvert,
|
|
.avxvnni,
|
|
.avxvnniint8,
|
|
.bmi,
|
|
.bmi2,
|
|
.cldemote,
|
|
.clflushopt,
|
|
.clwb,
|
|
.cmov,
|
|
.cmpccxadd,
|
|
.crc32,
|
|
.cx16,
|
|
.enqcmd,
|
|
.f16c,
|
|
.fast_movbe,
|
|
.fma,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.gfni,
|
|
.hreset,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.mmx,
|
|
.movbe,
|
|
.movdir64b,
|
|
.movdiri,
|
|
.no_bypass_delay,
|
|
.nopl,
|
|
.pconfig,
|
|
.pku,
|
|
.popcnt,
|
|
.prfchw,
|
|
.ptwrite,
|
|
.raoint,
|
|
.rdpid,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.serialize,
|
|
.sha,
|
|
.shstk,
|
|
.slow_incdec,
|
|
.slow_lea,
|
|
.slow_two_mem_ops,
|
|
.uintr,
|
|
.use_glm_div_sqrt_costs,
|
|
.vaes,
|
|
.vpclmulqdq,
|
|
.vzeroupper,
|
|
.waitpkg,
|
|
.widekl,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const graniterapids = CpuModel{
|
|
.name = "graniterapids",
|
|
.llvm_name = "graniterapids",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.allow_light_256_bit,
|
|
.amx_bf16,
|
|
.amx_fp16,
|
|
.amx_int8,
|
|
.avx512bf16,
|
|
.avx512bitalg,
|
|
.avx512cd,
|
|
.avx512fp16,
|
|
.avx512ifma,
|
|
.avx512vbmi,
|
|
.avx512vbmi2,
|
|
.avx512vnni,
|
|
.avx512vpopcntdq,
|
|
.avxvnni,
|
|
.bmi,
|
|
.bmi2,
|
|
.cldemote,
|
|
.clflushopt,
|
|
.clwb,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.enqcmd,
|
|
.ermsb,
|
|
.false_deps_getmant,
|
|
.false_deps_mulc,
|
|
.false_deps_mullq,
|
|
.false_deps_perm,
|
|
.false_deps_range,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fsgsbase,
|
|
.fsrm,
|
|
.fxsr,
|
|
.gfni,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.movdir64b,
|
|
.movdiri,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pconfig,
|
|
.pku,
|
|
.popcnt,
|
|
.prefer_256_bit,
|
|
.prefetchi,
|
|
.prfchw,
|
|
.ptwrite,
|
|
.rdpid,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.serialize,
|
|
.sha,
|
|
.shstk,
|
|
.tsxldtrk,
|
|
.tuning_fast_imm_vector_shift,
|
|
.uintr,
|
|
.vaes,
|
|
.vpclmulqdq,
|
|
.vzeroupper,
|
|
.waitpkg,
|
|
.wbnoinvd,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const graniterapids_d = CpuModel{
|
|
.name = "graniterapids_d",
|
|
.llvm_name = "graniterapids-d",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.allow_light_256_bit,
|
|
.amx_bf16,
|
|
.amx_complex,
|
|
.amx_fp16,
|
|
.amx_int8,
|
|
.avx512bf16,
|
|
.avx512bitalg,
|
|
.avx512cd,
|
|
.avx512fp16,
|
|
.avx512ifma,
|
|
.avx512vbmi,
|
|
.avx512vbmi2,
|
|
.avx512vnni,
|
|
.avx512vpopcntdq,
|
|
.avxvnni,
|
|
.bmi,
|
|
.bmi2,
|
|
.cldemote,
|
|
.clflushopt,
|
|
.clwb,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.enqcmd,
|
|
.ermsb,
|
|
.false_deps_getmant,
|
|
.false_deps_mulc,
|
|
.false_deps_mullq,
|
|
.false_deps_perm,
|
|
.false_deps_range,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fsgsbase,
|
|
.fsrm,
|
|
.fxsr,
|
|
.gfni,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.movdir64b,
|
|
.movdiri,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pconfig,
|
|
.pku,
|
|
.popcnt,
|
|
.prefer_256_bit,
|
|
.prefetchi,
|
|
.prfchw,
|
|
.ptwrite,
|
|
.rdpid,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.serialize,
|
|
.sha,
|
|
.shstk,
|
|
.tsxldtrk,
|
|
.tuning_fast_imm_vector_shift,
|
|
.uintr,
|
|
.vaes,
|
|
.vpclmulqdq,
|
|
.vzeroupper,
|
|
.waitpkg,
|
|
.wbnoinvd,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const haswell = CpuModel{
|
|
.name = "haswell",
|
|
.llvm_name = "haswell",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.allow_light_256_bit,
|
|
.avx2,
|
|
.bmi,
|
|
.bmi2,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.ermsb,
|
|
.f16c,
|
|
.false_deps_lzcnt_tzcnt,
|
|
.false_deps_popcnt,
|
|
.fast_15bytenop,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fma,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.rdrnd,
|
|
.sahf,
|
|
.slow_3ops_lea,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsaveopt,
|
|
}),
|
|
};
|
|
pub const @"i386" = CpuModel{
|
|
.name = "i386",
|
|
.llvm_name = "i386",
|
|
.features = featureSet(&[_]Feature{
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const @"i486" = CpuModel{
|
|
.name = "i486",
|
|
.llvm_name = "i486",
|
|
.features = featureSet(&[_]Feature{
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const @"i586" = CpuModel{
|
|
.name = "i586",
|
|
.llvm_name = "i586",
|
|
.features = featureSet(&[_]Feature{
|
|
.cx8,
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const @"i686" = CpuModel{
|
|
.name = "i686",
|
|
.llvm_name = "i686",
|
|
.features = featureSet(&[_]Feature{
|
|
.cmov,
|
|
.cx8,
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const icelake_client = CpuModel{
|
|
.name = "icelake_client",
|
|
.llvm_name = "icelake-client",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.allow_light_256_bit,
|
|
.avx512bitalg,
|
|
.avx512cd,
|
|
.avx512dq,
|
|
.avx512ifma,
|
|
.avx512vbmi,
|
|
.avx512vbmi2,
|
|
.avx512vl,
|
|
.avx512vnni,
|
|
.avx512vpopcntdq,
|
|
.bmi,
|
|
.bmi2,
|
|
.clflushopt,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.ermsb,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fsgsbase,
|
|
.fsrm,
|
|
.fxsr,
|
|
.gfni,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pku,
|
|
.popcnt,
|
|
.prefer_256_bit,
|
|
.prfchw,
|
|
.rdpid,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.sha,
|
|
.tuning_fast_imm_vector_shift,
|
|
.vaes,
|
|
.vpclmulqdq,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const icelake_server = CpuModel{
|
|
.name = "icelake_server",
|
|
.llvm_name = "icelake-server",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.allow_light_256_bit,
|
|
.avx512bitalg,
|
|
.avx512cd,
|
|
.avx512dq,
|
|
.avx512ifma,
|
|
.avx512vbmi,
|
|
.avx512vbmi2,
|
|
.avx512vl,
|
|
.avx512vnni,
|
|
.avx512vpopcntdq,
|
|
.bmi,
|
|
.bmi2,
|
|
.clflushopt,
|
|
.clwb,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.ermsb,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fsgsbase,
|
|
.fsrm,
|
|
.fxsr,
|
|
.gfni,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pconfig,
|
|
.pku,
|
|
.popcnt,
|
|
.prefer_256_bit,
|
|
.prfchw,
|
|
.rdpid,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.sha,
|
|
.tuning_fast_imm_vector_shift,
|
|
.vaes,
|
|
.vpclmulqdq,
|
|
.vzeroupper,
|
|
.wbnoinvd,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const ivybridge = CpuModel{
|
|
.name = "ivybridge",
|
|
.llvm_name = "ivybridge",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.f16c,
|
|
.false_deps_popcnt,
|
|
.fast_15bytenop,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.macrofusion,
|
|
.mmx,
|
|
.no_bypass_delay_mov,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.rdrnd,
|
|
.sahf,
|
|
.slow_3ops_lea,
|
|
.slow_unaligned_mem_32,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsaveopt,
|
|
}),
|
|
};
|
|
pub const k6 = CpuModel{
|
|
.name = "k6",
|
|
.llvm_name = "k6",
|
|
.features = featureSet(&[_]Feature{
|
|
.cx8,
|
|
.mmx,
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const k6_2 = CpuModel{
|
|
.name = "k6_2",
|
|
.llvm_name = "k6-2",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnow",
|
|
.cx8,
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const k6_3 = CpuModel{
|
|
.name = "k6_3",
|
|
.llvm_name = "k6-3",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnow",
|
|
.cx8,
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const k8 = CpuModel{
|
|
.name = "k8",
|
|
.llvm_name = "k8",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnowa",
|
|
.@"64bit",
|
|
.cmov,
|
|
.cx8,
|
|
.fast_scalar_shift_masks,
|
|
.fxsr,
|
|
.nopl,
|
|
.sbb_dep_breaking,
|
|
.slow_shld,
|
|
.slow_unaligned_mem_16,
|
|
.sse2,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const k8_sse3 = CpuModel{
|
|
.name = "k8_sse3",
|
|
.llvm_name = "k8-sse3",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnowa",
|
|
.@"64bit",
|
|
.cmov,
|
|
.cx16,
|
|
.fast_scalar_shift_masks,
|
|
.fxsr,
|
|
.nopl,
|
|
.sbb_dep_breaking,
|
|
.slow_shld,
|
|
.slow_unaligned_mem_16,
|
|
.sse3,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const knl = CpuModel{
|
|
.name = "knl",
|
|
.llvm_name = "knl",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.aes,
|
|
.avx512cd,
|
|
.avx512er,
|
|
.avx512pf,
|
|
.bmi,
|
|
.bmi2,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.fast_gather,
|
|
.fast_movbe,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.lzcnt,
|
|
.mmx,
|
|
.movbe,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prefer_mask_registers,
|
|
.prefetchwt1,
|
|
.prfchw,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.slow_3ops_lea,
|
|
.slow_incdec,
|
|
.slow_pmaddwd,
|
|
.slow_two_mem_ops,
|
|
.x87,
|
|
.xsaveopt,
|
|
}),
|
|
};
|
|
pub const knm = CpuModel{
|
|
.name = "knm",
|
|
.llvm_name = "knm",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.aes,
|
|
.avx512cd,
|
|
.avx512er,
|
|
.avx512pf,
|
|
.avx512vpopcntdq,
|
|
.bmi,
|
|
.bmi2,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.fast_gather,
|
|
.fast_movbe,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.lzcnt,
|
|
.mmx,
|
|
.movbe,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prefer_mask_registers,
|
|
.prefetchwt1,
|
|
.prfchw,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.slow_3ops_lea,
|
|
.slow_incdec,
|
|
.slow_pmaddwd,
|
|
.slow_two_mem_ops,
|
|
.x87,
|
|
.xsaveopt,
|
|
}),
|
|
};
|
|
pub const lakemont = CpuModel{
|
|
.name = "lakemont",
|
|
.llvm_name = "lakemont",
|
|
.features = featureSet(&[_]Feature{
|
|
.cx8,
|
|
.slow_unaligned_mem_16,
|
|
.soft_float,
|
|
.vzeroupper,
|
|
}),
|
|
};
|
|
pub const meteorlake = CpuModel{
|
|
.name = "meteorlake",
|
|
.llvm_name = "meteorlake",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.allow_light_256_bit,
|
|
.avxvnni,
|
|
.bmi,
|
|
.bmi2,
|
|
.cldemote,
|
|
.clflushopt,
|
|
.clwb,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.f16c,
|
|
.false_deps_perm,
|
|
.false_deps_popcnt,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fma,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.gfni,
|
|
.hreset,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.movdir64b,
|
|
.movdiri,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pconfig,
|
|
.pku,
|
|
.popcnt,
|
|
.prefer_movmsk_over_vtest,
|
|
.prfchw,
|
|
.ptwrite,
|
|
.rdpid,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.serialize,
|
|
.sha,
|
|
.shstk,
|
|
.slow_3ops_lea,
|
|
.tuning_fast_imm_vector_shift,
|
|
.vaes,
|
|
.vpclmulqdq,
|
|
.vzeroupper,
|
|
.waitpkg,
|
|
.widekl,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const nehalem = CpuModel{
|
|
.name = "nehalem",
|
|
.llvm_name = "nehalem",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.fxsr,
|
|
.macrofusion,
|
|
.mmx,
|
|
.no_bypass_delay_mov,
|
|
.nopl,
|
|
.popcnt,
|
|
.sahf,
|
|
.sse4_2,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const nocona = CpuModel{
|
|
.name = "nocona",
|
|
.llvm_name = "nocona",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.cmov,
|
|
.cx16,
|
|
.fxsr,
|
|
.mmx,
|
|
.nopl,
|
|
.slow_unaligned_mem_16,
|
|
.sse3,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const opteron = CpuModel{
|
|
.name = "opteron",
|
|
.llvm_name = "opteron",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnowa",
|
|
.@"64bit",
|
|
.cmov,
|
|
.cx8,
|
|
.fast_scalar_shift_masks,
|
|
.fxsr,
|
|
.nopl,
|
|
.sbb_dep_breaking,
|
|
.slow_shld,
|
|
.slow_unaligned_mem_16,
|
|
.sse2,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const opteron_sse3 = CpuModel{
|
|
.name = "opteron_sse3",
|
|
.llvm_name = "opteron-sse3",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnowa",
|
|
.@"64bit",
|
|
.cmov,
|
|
.cx16,
|
|
.fast_scalar_shift_masks,
|
|
.fxsr,
|
|
.nopl,
|
|
.sbb_dep_breaking,
|
|
.slow_shld,
|
|
.slow_unaligned_mem_16,
|
|
.sse3,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const penryn = CpuModel{
|
|
.name = "penryn",
|
|
.llvm_name = "penryn",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.cmov,
|
|
.cx16,
|
|
.fxsr,
|
|
.macrofusion,
|
|
.mmx,
|
|
.nopl,
|
|
.sahf,
|
|
.slow_unaligned_mem_16,
|
|
.sse4_1,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const pentium = CpuModel{
|
|
.name = "pentium",
|
|
.llvm_name = "pentium",
|
|
.features = featureSet(&[_]Feature{
|
|
.cx8,
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const pentium2 = CpuModel{
|
|
.name = "pentium2",
|
|
.llvm_name = "pentium2",
|
|
.features = featureSet(&[_]Feature{
|
|
.cmov,
|
|
.cx8,
|
|
.fxsr,
|
|
.mmx,
|
|
.nopl,
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const pentium3 = CpuModel{
|
|
.name = "pentium3",
|
|
.llvm_name = "pentium3",
|
|
.features = featureSet(&[_]Feature{
|
|
.cmov,
|
|
.cx8,
|
|
.fxsr,
|
|
.mmx,
|
|
.nopl,
|
|
.slow_unaligned_mem_16,
|
|
.sse,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const pentium3m = CpuModel{
|
|
.name = "pentium3m",
|
|
.llvm_name = "pentium3m",
|
|
.features = featureSet(&[_]Feature{
|
|
.cmov,
|
|
.cx8,
|
|
.fxsr,
|
|
.mmx,
|
|
.nopl,
|
|
.slow_unaligned_mem_16,
|
|
.sse,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const pentium4 = CpuModel{
|
|
.name = "pentium4",
|
|
.llvm_name = "pentium4",
|
|
.features = featureSet(&[_]Feature{
|
|
.cmov,
|
|
.cx8,
|
|
.fxsr,
|
|
.mmx,
|
|
.nopl,
|
|
.slow_unaligned_mem_16,
|
|
.sse2,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const pentium_m = CpuModel{
|
|
.name = "pentium_m",
|
|
.llvm_name = "pentium-m",
|
|
.features = featureSet(&[_]Feature{
|
|
.cmov,
|
|
.cx8,
|
|
.fxsr,
|
|
.mmx,
|
|
.nopl,
|
|
.slow_unaligned_mem_16,
|
|
.sse2,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const pentium_mmx = CpuModel{
|
|
.name = "pentium_mmx",
|
|
.llvm_name = "pentium-mmx",
|
|
.features = featureSet(&[_]Feature{
|
|
.cx8,
|
|
.mmx,
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const pentiumpro = CpuModel{
|
|
.name = "pentiumpro",
|
|
.llvm_name = "pentiumpro",
|
|
.features = featureSet(&[_]Feature{
|
|
.cmov,
|
|
.cx8,
|
|
.nopl,
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const prescott = CpuModel{
|
|
.name = "prescott",
|
|
.llvm_name = "prescott",
|
|
.features = featureSet(&[_]Feature{
|
|
.cmov,
|
|
.cx8,
|
|
.fxsr,
|
|
.mmx,
|
|
.nopl,
|
|
.slow_unaligned_mem_16,
|
|
.sse3,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const raptorlake = CpuModel{
|
|
.name = "raptorlake",
|
|
.llvm_name = "raptorlake",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.allow_light_256_bit,
|
|
.avxvnni,
|
|
.bmi,
|
|
.bmi2,
|
|
.cldemote,
|
|
.clflushopt,
|
|
.clwb,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.f16c,
|
|
.false_deps_perm,
|
|
.false_deps_popcnt,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fma,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.gfni,
|
|
.hreset,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.movdir64b,
|
|
.movdiri,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pconfig,
|
|
.pku,
|
|
.popcnt,
|
|
.prefer_movmsk_over_vtest,
|
|
.prfchw,
|
|
.ptwrite,
|
|
.rdpid,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.serialize,
|
|
.sha,
|
|
.shstk,
|
|
.slow_3ops_lea,
|
|
.tuning_fast_imm_vector_shift,
|
|
.vaes,
|
|
.vpclmulqdq,
|
|
.vzeroupper,
|
|
.waitpkg,
|
|
.widekl,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const rocketlake = CpuModel{
|
|
.name = "rocketlake",
|
|
.llvm_name = "rocketlake",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.allow_light_256_bit,
|
|
.avx512bitalg,
|
|
.avx512cd,
|
|
.avx512dq,
|
|
.avx512ifma,
|
|
.avx512vbmi,
|
|
.avx512vbmi2,
|
|
.avx512vl,
|
|
.avx512vnni,
|
|
.avx512vpopcntdq,
|
|
.bmi,
|
|
.bmi2,
|
|
.clflushopt,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.ermsb,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fsgsbase,
|
|
.fsrm,
|
|
.fxsr,
|
|
.gfni,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pku,
|
|
.popcnt,
|
|
.prefer_256_bit,
|
|
.prfchw,
|
|
.rdpid,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.sha,
|
|
.tuning_fast_imm_vector_shift,
|
|
.vaes,
|
|
.vpclmulqdq,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const sandybridge = CpuModel{
|
|
.name = "sandybridge",
|
|
.llvm_name = "sandybridge",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.avx,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.false_deps_popcnt,
|
|
.fast_15bytenop,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.macrofusion,
|
|
.mmx,
|
|
.no_bypass_delay_mov,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.sahf,
|
|
.slow_3ops_lea,
|
|
.slow_unaligned_mem_32,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsaveopt,
|
|
}),
|
|
};
|
|
pub const sapphirerapids = CpuModel{
|
|
.name = "sapphirerapids",
|
|
.llvm_name = "sapphirerapids",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.allow_light_256_bit,
|
|
.amx_bf16,
|
|
.amx_int8,
|
|
.avx512bf16,
|
|
.avx512bitalg,
|
|
.avx512cd,
|
|
.avx512fp16,
|
|
.avx512ifma,
|
|
.avx512vbmi,
|
|
.avx512vbmi2,
|
|
.avx512vnni,
|
|
.avx512vpopcntdq,
|
|
.avxvnni,
|
|
.bmi,
|
|
.bmi2,
|
|
.cldemote,
|
|
.clflushopt,
|
|
.clwb,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.enqcmd,
|
|
.ermsb,
|
|
.false_deps_getmant,
|
|
.false_deps_mulc,
|
|
.false_deps_mullq,
|
|
.false_deps_perm,
|
|
.false_deps_range,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fsgsbase,
|
|
.fsrm,
|
|
.fxsr,
|
|
.gfni,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.movdir64b,
|
|
.movdiri,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pconfig,
|
|
.pku,
|
|
.popcnt,
|
|
.prefer_256_bit,
|
|
.prfchw,
|
|
.ptwrite,
|
|
.rdpid,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.serialize,
|
|
.sha,
|
|
.shstk,
|
|
.tsxldtrk,
|
|
.tuning_fast_imm_vector_shift,
|
|
.uintr,
|
|
.vaes,
|
|
.vpclmulqdq,
|
|
.vzeroupper,
|
|
.waitpkg,
|
|
.wbnoinvd,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const sierraforest = CpuModel{
|
|
.name = "sierraforest",
|
|
.llvm_name = "sierraforest",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.avxifma,
|
|
.avxneconvert,
|
|
.avxvnni,
|
|
.avxvnniint8,
|
|
.bmi,
|
|
.bmi2,
|
|
.cldemote,
|
|
.clflushopt,
|
|
.clwb,
|
|
.cmov,
|
|
.cmpccxadd,
|
|
.crc32,
|
|
.cx16,
|
|
.enqcmd,
|
|
.f16c,
|
|
.fast_movbe,
|
|
.fma,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.gfni,
|
|
.hreset,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.mmx,
|
|
.movbe,
|
|
.movdir64b,
|
|
.movdiri,
|
|
.no_bypass_delay,
|
|
.nopl,
|
|
.pconfig,
|
|
.pku,
|
|
.popcnt,
|
|
.prfchw,
|
|
.ptwrite,
|
|
.rdpid,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.serialize,
|
|
.sha,
|
|
.shstk,
|
|
.slow_incdec,
|
|
.slow_lea,
|
|
.slow_two_mem_ops,
|
|
.uintr,
|
|
.use_glm_div_sqrt_costs,
|
|
.vaes,
|
|
.vpclmulqdq,
|
|
.vzeroupper,
|
|
.waitpkg,
|
|
.widekl,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const silvermont = CpuModel{
|
|
.name = "silvermont",
|
|
.llvm_name = "silvermont",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.false_deps_popcnt,
|
|
.fast_7bytenop,
|
|
.fast_movbe,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prfchw,
|
|
.rdrnd,
|
|
.sahf,
|
|
.slow_incdec,
|
|
.slow_lea,
|
|
.slow_pmulld,
|
|
.slow_two_mem_ops,
|
|
.sse4_2,
|
|
.use_slm_arith_costs,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const skx = CpuModel{
|
|
.name = "skx",
|
|
.llvm_name = "skx",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.aes,
|
|
.allow_light_256_bit,
|
|
.avx512bw,
|
|
.avx512cd,
|
|
.avx512dq,
|
|
.avx512vl,
|
|
.bmi,
|
|
.bmi2,
|
|
.clflushopt,
|
|
.clwb,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.ermsb,
|
|
.false_deps_popcnt,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.faster_shift_than_shuffle,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pclmul,
|
|
.pku,
|
|
.popcnt,
|
|
.prefer_256_bit,
|
|
.prfchw,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.slow_3ops_lea,
|
|
.tuning_fast_imm_vector_shift,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const skylake = CpuModel{
|
|
.name = "skylake",
|
|
.llvm_name = "skylake",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.aes,
|
|
.allow_light_256_bit,
|
|
.avx2,
|
|
.bmi,
|
|
.bmi2,
|
|
.clflushopt,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.ermsb,
|
|
.f16c,
|
|
.false_deps_popcnt,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fma,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prfchw,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.slow_3ops_lea,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const skylake_avx512 = CpuModel{
|
|
.name = "skylake_avx512",
|
|
.llvm_name = "skylake-avx512",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.aes,
|
|
.allow_light_256_bit,
|
|
.avx512bw,
|
|
.avx512cd,
|
|
.avx512dq,
|
|
.avx512vl,
|
|
.bmi,
|
|
.bmi2,
|
|
.clflushopt,
|
|
.clwb,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.ermsb,
|
|
.false_deps_popcnt,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.faster_shift_than_shuffle,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pclmul,
|
|
.pku,
|
|
.popcnt,
|
|
.prefer_256_bit,
|
|
.prfchw,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.slow_3ops_lea,
|
|
.tuning_fast_imm_vector_shift,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const slm = CpuModel{
|
|
.name = "slm",
|
|
.llvm_name = "slm",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.false_deps_popcnt,
|
|
.fast_7bytenop,
|
|
.fast_movbe,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prfchw,
|
|
.rdrnd,
|
|
.sahf,
|
|
.slow_incdec,
|
|
.slow_lea,
|
|
.slow_pmulld,
|
|
.slow_two_mem_ops,
|
|
.sse4_2,
|
|
.use_slm_arith_costs,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const tigerlake = CpuModel{
|
|
.name = "tigerlake",
|
|
.llvm_name = "tigerlake",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.allow_light_256_bit,
|
|
.avx512bitalg,
|
|
.avx512cd,
|
|
.avx512dq,
|
|
.avx512ifma,
|
|
.avx512vbmi,
|
|
.avx512vbmi2,
|
|
.avx512vl,
|
|
.avx512vnni,
|
|
.avx512vp2intersect,
|
|
.avx512vpopcntdq,
|
|
.bmi,
|
|
.bmi2,
|
|
.clflushopt,
|
|
.clwb,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.ermsb,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fsgsbase,
|
|
.fsrm,
|
|
.fxsr,
|
|
.gfni,
|
|
.idivq_to_divl,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.movdir64b,
|
|
.movdiri,
|
|
.no_bypass_delay_blend,
|
|
.no_bypass_delay_mov,
|
|
.no_bypass_delay_shuffle,
|
|
.nopl,
|
|
.pku,
|
|
.popcnt,
|
|
.prefer_256_bit,
|
|
.prfchw,
|
|
.rdpid,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.sha,
|
|
.shstk,
|
|
.tuning_fast_imm_vector_shift,
|
|
.vaes,
|
|
.vpclmulqdq,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const tremont = CpuModel{
|
|
.name = "tremont",
|
|
.llvm_name = "tremont",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.aes,
|
|
.clflushopt,
|
|
.clwb,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.fast_movbe,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.gfni,
|
|
.mmx,
|
|
.movbe,
|
|
.no_bypass_delay,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prfchw,
|
|
.ptwrite,
|
|
.rdpid,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.sha,
|
|
.slow_incdec,
|
|
.slow_lea,
|
|
.slow_two_mem_ops,
|
|
.sse4_2,
|
|
.use_glm_div_sqrt_costs,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const westmere = CpuModel{
|
|
.name = "westmere",
|
|
.llvm_name = "westmere",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.fxsr,
|
|
.macrofusion,
|
|
.mmx,
|
|
.no_bypass_delay_mov,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.sahf,
|
|
.sse4_2,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const winchip2 = CpuModel{
|
|
.name = "winchip2",
|
|
.llvm_name = "winchip2",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"3dnow",
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const winchip_c6 = CpuModel{
|
|
.name = "winchip_c6",
|
|
.llvm_name = "winchip-c6",
|
|
.features = featureSet(&[_]Feature{
|
|
.mmx,
|
|
.slow_unaligned_mem_16,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const x86_64 = CpuModel{
|
|
.name = "x86_64",
|
|
.llvm_name = "x86-64",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.cmov,
|
|
.cx8,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.macrofusion,
|
|
.mmx,
|
|
.nopl,
|
|
.slow_3ops_lea,
|
|
.slow_incdec,
|
|
.sse2,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const x86_64_v2 = CpuModel{
|
|
.name = "x86_64_v2",
|
|
.llvm_name = "x86-64-v2",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.false_deps_popcnt,
|
|
.fast_15bytenop,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.macrofusion,
|
|
.mmx,
|
|
.nopl,
|
|
.popcnt,
|
|
.sahf,
|
|
.slow_3ops_lea,
|
|
.slow_unaligned_mem_32,
|
|
.sse4_2,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const x86_64_v3 = CpuModel{
|
|
.name = "x86_64_v3",
|
|
.llvm_name = "x86-64-v3",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.allow_light_256_bit,
|
|
.avx2,
|
|
.bmi,
|
|
.bmi2,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.f16c,
|
|
.false_deps_lzcnt_tzcnt,
|
|
.false_deps_popcnt,
|
|
.fast_15bytenop,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fma,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.nopl,
|
|
.popcnt,
|
|
.sahf,
|
|
.slow_3ops_lea,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsave,
|
|
}),
|
|
};
|
|
pub const x86_64_v4 = CpuModel{
|
|
.name = "x86_64_v4",
|
|
.llvm_name = "x86-64-v4",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.allow_light_256_bit,
|
|
.avx512bw,
|
|
.avx512cd,
|
|
.avx512dq,
|
|
.avx512vl,
|
|
.bmi,
|
|
.bmi2,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.false_deps_popcnt,
|
|
.fast_15bytenop,
|
|
.fast_gather,
|
|
.fast_scalar_fsqrt,
|
|
.fast_shld_rotate,
|
|
.fast_variable_crosslane_shuffle,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fxsr,
|
|
.idivq_to_divl,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.nopl,
|
|
.popcnt,
|
|
.prefer_256_bit,
|
|
.sahf,
|
|
.slow_3ops_lea,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsave,
|
|
}),
|
|
};
|
|
pub const yonah = CpuModel{
|
|
.name = "yonah",
|
|
.llvm_name = "yonah",
|
|
.features = featureSet(&[_]Feature{
|
|
.cmov,
|
|
.cx8,
|
|
.fxsr,
|
|
.mmx,
|
|
.nopl,
|
|
.slow_unaligned_mem_16,
|
|
.sse3,
|
|
.vzeroupper,
|
|
.x87,
|
|
}),
|
|
};
|
|
pub const znver1 = CpuModel{
|
|
.name = "znver1",
|
|
.llvm_name = "znver1",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.aes,
|
|
.allow_light_256_bit,
|
|
.avx2,
|
|
.bmi,
|
|
.bmi2,
|
|
.branchfusion,
|
|
.clflushopt,
|
|
.clzero,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.f16c,
|
|
.fast_15bytenop,
|
|
.fast_bextr,
|
|
.fast_lzcnt,
|
|
.fast_movbe,
|
|
.fast_scalar_fsqrt,
|
|
.fast_scalar_shift_masks,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fma,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.lzcnt,
|
|
.mmx,
|
|
.movbe,
|
|
.mwaitx,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prfchw,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.sbb_dep_breaking,
|
|
.sha,
|
|
.slow_shld,
|
|
.sse4a,
|
|
.vzeroupper,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const znver2 = CpuModel{
|
|
.name = "znver2",
|
|
.llvm_name = "znver2",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.aes,
|
|
.allow_light_256_bit,
|
|
.avx2,
|
|
.bmi,
|
|
.bmi2,
|
|
.branchfusion,
|
|
.clflushopt,
|
|
.clwb,
|
|
.clzero,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.f16c,
|
|
.fast_15bytenop,
|
|
.fast_bextr,
|
|
.fast_lzcnt,
|
|
.fast_movbe,
|
|
.fast_scalar_fsqrt,
|
|
.fast_scalar_shift_masks,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fma,
|
|
.fsgsbase,
|
|
.fxsr,
|
|
.lzcnt,
|
|
.mmx,
|
|
.movbe,
|
|
.mwaitx,
|
|
.nopl,
|
|
.pclmul,
|
|
.popcnt,
|
|
.prfchw,
|
|
.rdpid,
|
|
.rdpru,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.sbb_dep_breaking,
|
|
.sha,
|
|
.slow_shld,
|
|
.sse4a,
|
|
.vzeroupper,
|
|
.wbnoinvd,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const znver3 = CpuModel{
|
|
.name = "znver3",
|
|
.llvm_name = "znver3",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.allow_light_256_bit,
|
|
.avx2,
|
|
.bmi,
|
|
.bmi2,
|
|
.branchfusion,
|
|
.clflushopt,
|
|
.clwb,
|
|
.clzero,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.f16c,
|
|
.fast_15bytenop,
|
|
.fast_bextr,
|
|
.fast_lzcnt,
|
|
.fast_movbe,
|
|
.fast_scalar_fsqrt,
|
|
.fast_scalar_shift_masks,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fma,
|
|
.fsgsbase,
|
|
.fsrm,
|
|
.fxsr,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.mwaitx,
|
|
.nopl,
|
|
.pku,
|
|
.popcnt,
|
|
.prfchw,
|
|
.rdpid,
|
|
.rdpru,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.sbb_dep_breaking,
|
|
.sha,
|
|
.slow_shld,
|
|
.sse4a,
|
|
.vaes,
|
|
.vpclmulqdq,
|
|
.vzeroupper,
|
|
.wbnoinvd,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
pub const znver4 = CpuModel{
|
|
.name = "znver4",
|
|
.llvm_name = "znver4",
|
|
.features = featureSet(&[_]Feature{
|
|
.@"64bit",
|
|
.adx,
|
|
.allow_light_256_bit,
|
|
.avx512bf16,
|
|
.avx512bitalg,
|
|
.avx512cd,
|
|
.avx512dq,
|
|
.avx512ifma,
|
|
.avx512vbmi,
|
|
.avx512vbmi2,
|
|
.avx512vl,
|
|
.avx512vnni,
|
|
.avx512vpopcntdq,
|
|
.bmi,
|
|
.bmi2,
|
|
.branchfusion,
|
|
.clflushopt,
|
|
.clwb,
|
|
.clzero,
|
|
.cmov,
|
|
.crc32,
|
|
.cx16,
|
|
.fast_15bytenop,
|
|
.fast_bextr,
|
|
.fast_lzcnt,
|
|
.fast_movbe,
|
|
.fast_scalar_fsqrt,
|
|
.fast_scalar_shift_masks,
|
|
.fast_variable_perlane_shuffle,
|
|
.fast_vector_fsqrt,
|
|
.fsgsbase,
|
|
.fsrm,
|
|
.fxsr,
|
|
.gfni,
|
|
.invpcid,
|
|
.lzcnt,
|
|
.macrofusion,
|
|
.mmx,
|
|
.movbe,
|
|
.mwaitx,
|
|
.nopl,
|
|
.pku,
|
|
.popcnt,
|
|
.prfchw,
|
|
.rdpid,
|
|
.rdpru,
|
|
.rdrnd,
|
|
.rdseed,
|
|
.sahf,
|
|
.sbb_dep_breaking,
|
|
.sha,
|
|
.shstk,
|
|
.slow_shld,
|
|
.sse4a,
|
|
.vaes,
|
|
.vpclmulqdq,
|
|
.vzeroupper,
|
|
.wbnoinvd,
|
|
.x87,
|
|
.xsavec,
|
|
.xsaveopt,
|
|
.xsaves,
|
|
}),
|
|
};
|
|
};
|